Aurascape AI Security Platform

Aurascape, Inc. is an AI security company based in Santa Clara, California, founded by Moinul Khan (CEO), Patrick Xu (CTO), Viswesh and Liang Li, veterans of Palo Alto Networks, Zscaler, FireEye, Juniper and Google. It came out of a year in stealth in April 2025 with 50 million USD from Mayfield Fund and Menlo Ventures, to help enterprises adopt generative AI without losing visibility or leaking data.

The platform inspects AI traffic inline across browsers, desktop apps, CLI tools, SaaS apps with embedded AI, IDEs and agent workflows. It discovers and risk-scores tens of thousands of AI apps, decodes prompts, responses, identity and intent, and enforces policy by role and conversation context. Data protection uses classification for hundreds of data types, fingerprinting and masking. For teams that build agents, it maps MCP servers and tool connections, runs adversarial tests, applies runtime guardrails and governs tool calls through its Zero-Bypass MCP Gateway.

• AI discovery and risk scoring: Detects shadow and embedded AI apps, copilots and agents and scores their risk in real time.
• Deep intention decoders: Analyse prompts, responses, user identity and intent to apply conversation-based conditional access.
• AI-driven data protection: Classifies hundreds of data types, fingerprints sensitive content and masks values before they reach the model.
• Copilot readiness: Finds overshared files, tags sensitive content and supports unlearning for copilots such as Microsoft Copilot.
• Agent build security: Maps MCP servers and tool connections, runs adversarial tests and detects code path and CVE risks.
• Zero-Bypass MCP Gateway: Combines an MCP gateway and AI proxy with tool signing to govern agent tool calls in production.

A Vietnamese bank preparing a Microsoft Copilot rollout can find overshared documents first and then mask account numbers in prompts. A software outsourcing company can control which coding assistants and models developers use and keep client source code out of personal AI accounts. A retailer building customer service agents can route all MCP tool calls through a gateway and test the agents for prompt injection before launch.

Security teams see every AI app, copilot, agent and model in use, including AI embedded in SaaS apps they already trust, and can tell enterprise from personal accounts. Sensitive values can be masked before they reach the model and restored in the answer, so staff keep working instead of being blocked. Model-level policy lets a company approve a platform such as Amazon Bedrock while restricting specific models inside it. One platform covers both AI use by employees and AI agents built in-house, reducing the number of separate tools.

AI features are being embedded into SaaS tools and employees often pay for their own GenAI apps, so URL-level controls in firewalls and SSE no longer show what data flows through prompts and agent tool calls. As companies move agents into production, MCP traffic becomes a new control point. Vietnamese and Southeast Asian enterprises rolling out Copilot and ChatGPT under personal data protection rules need intent-aware visibility and data masking for this traffic.

Aurascape, Inc. is located at 2625 Augustine Drive, Santa Clara, California. It launched from stealth on 8 April 2025 with 50 million USD led by Mayfield Fund and Menlo Ventures, with Celesta Capital and individual investors including former Palo Alto Networks CEO Mark McLaughlin, former Symantec CEO Greg Clark and Lip-Bu Tan. Advisors include CISOs of AT&T and Mastercard and the CIO of Databricks.